2026-09-04

Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills

👍 496

09/02 08:00

Autonomous agents are beginning to carry out machine-learning (ML) research end to end. These agents combine a model backbone with a harness for planning, execution, memory, and verification, but this architecture still leaves domain-specific know-how outside the agent. We call this missing layer op

中文介绍 面向AI4AI技能提取，提出Repo-To-Skill方法，将GitHub仓库蒸馏为可复用的技能库，弥补智能体在领域特定知识上的缺失。其核心技术在于自动化提取与组织技能，以提升自主机器学习研究的执行能力。

HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?

👍 225

09/01 08:00

As agents move from research prototypes to deployed tools, their capability increasingly depends on model-external execution infrastructure, commonly termed the agent harness. Changing this harness while holding model weights fixed can substantially alter task performance. Current agent evaluations

中文介绍 研究LLM能否自主创建和进化智能体运行环境（harness）。提出HarnessDev框架，让模型在固定权重下迭代优化harness代码，从而显著提升任务性能，突破仅优化模型参数的瓶颈。

Aspire: Can Models Self-Evolve from Vague Goals?

👍 197

08/31 08:00

Many important forms of human learning begin with a vague goal, such as "become a better physicist" or "improve at research." Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evo

中文介绍 针对模糊目标下的自我进化问题，提出Aspire方法，使LLM能解读模糊目标、识别能力缺口、决定学习方式并评估进步，摆脱对预设明确任务和外部奖励的依赖，实现更接近人类的自主学习。

EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction

👍 110

09/02 08:00

Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an agentic benchmark can cost hundreds to thousands of dollars, a price paid repeatedly across iterative development cycles. Prior efforts, centered on b

中文介绍 提出EarlyEval，通过早期结果预测降低智能体评估成本。基于部分轨迹提前预测最终结果，避免完整执行昂贵基准测试，在保持评估准确性的同时大幅节省计算开销，加速迭代开发。

It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning

👍 64

09/01 08:00

Retrieval is the first stage of modern search and advertising systems, selecting a candidate set from a large item universe for downstream ranking and auction. Recent work increasingly leverages LLMs to improve retrieval through query expansion, data synthesis, and retrieval-feedback training. Howev

中文介绍 提出生成式检索器与强化学习协同进化方法，解决传统检索中查询与文档匹配不足的问题。通过双智能体共同进化，利用LLM改进查询扩展和数据合成，使检索器更适应复杂信息需求。

Language Models Can Control Their Own Attention

👍 51

09/02 08:00

Language models spend most of their attention on a small fraction of context, yet they read the entire KV cache to find the few tokens that matter. If the user asks about a previous detail in a 1M-token conversation, global attention layers must scan the full context to generate each token of the re

中文介绍 发现语言模型注意力集中于少数token，却需扫描全部上下文。据此提出新方法让模型自主控制注意力，只处理相关token，显著减少KV cache访问，提升长上下文推理效率。

From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix

👍 33

09/01 08:00

Data-residency constraints force enterprises to self-host LLMs, but continuous adoption of newer models without decommissioning their predecessors expands the serving fleet, fragmenting a finite GPU pool. We consolidate traffic from over 200 internal applications onto a single model by closing quali

中文介绍 针对企业数据驻留限制，提出自托管LLM整合方案。通过后训练质量差距关闭，将200多个内部应用的流量整合到单一模型，解决GPU资源碎片化问题，提升推理效率。

On the Design Fundamentals of Pixel Text Representation Learning

👍 28

09/01 08:00

Text-rich visual inputs require models that can read, retrieve, and compress language directly in pixel space, yet existing pixel-text encoders struggle with fixed resolution pretraining, visual shortcut learning, weak visual grounding, and multilingual visual text understanding. In this work, we in

中文介绍 研究像素级文本表示学习的设计基础，旨在提升模型在像素空间中直接阅读、检索和压缩文本的能力。通过改进预训练架构和目标，解决固定分辨率、视觉捷径学习等挑战。

Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

👍 26

08/19 08:00

Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employ CLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semanti

中文介绍 针对知识型视觉问答的实体对齐检索问题，提出超越视觉相似度的方法。核心是构建实体级语义对齐，而非仅依赖CLIP的表面相似度，从而提升长尾实体问题的知识检索准确性。

S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

👍 26

08/31 08:00

Large language models (LLMs) increasingly interact with external environments and accumulate substantial behavioral experience, yet existing agent benchmarks largely evaluate them as fixed policies. It therefore remains unclear whether an agent can actively test its behavior, judge the resulting exp

中文介绍 提出S3Gym环境，让LLM在交互中主动测试行为、评判结果并自我改进。通过自我测试和评判循环，突破固定策略评估的局限，验证智能体持续自我提升的可行性。

ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes

👍 20

09/01 08:00

Compact token sequences are essential for efficient 3D generation. However, existing 3D tokenizers typically organize latent representations either over spatial regions or as fixed-size sets of global tokens, both suffering sharp reconstruction degradation when compressed to extremely low token budg

中文介绍 提出ZipTok3D，一种高保真3D tokenizer，通过紧凑token前缀实现低token预算下的高质量3D生成。克服现有方法在极低压缩率下重建质量严重下降的问题，提升生成效率。

AgentJudgeBench: A Multi-Difficulty Benchmark for Evaluating LLM Judges on Agentic Tool-Calling

👍 19

08/27 08:00

LLM judges are widely used to evaluate agentic tool-calling systems, yet their reliability on structured, dependency-driven workflows remains largely unexamined. We present AgentJudgeBench, the first benchmark to systematically study LLM-as-a-judge reliability for agentic tool-calling over workflow

中文介绍 介绍AgentJudgeBench，首个系统研究LLM作为裁判在代理工具调用中可靠性的基准。覆盖多难度工作流，发现现有评判模型在依赖驱动任务上的不足，并构建更难评估集以推动改进。

DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory

👍 18

09/01 08:00

Self-play is an effective paradigm for language-model self-evolution, but without guidance, solver performance can plateau or decline across rounds. Unguided methods steer question generation with signals such as difficulty, learnability, or diversity. These signals keep questions challenging and va

中文介绍 提出DiagEvo，利用分层错误记忆指导问题生成，引导自进化。通过记录并分析历史错误，使问题生成更具针对性和多样性，避免无引导自博弈导致的性能停滞或下降。

Cliff: Learning Process Rewards from the First Mistake

👍 15

09/02 08:00

Reinforcement learning with verifiable rewards (RLVR) has emerged as a powerful paradigm for large language model (LLM) post-training, but its reliance on coarse outcome rewards leads to limited guidance on intermediate reasoning processes. Existing approaches such as process reward modeling and on-

中文介绍 提出Cliff方法，从首个错误学习过程奖励，弥补结果奖励粗糙和过程奖励建模的不足。通过识别推理中的早期错误，提供更细粒度的训练信号，提升RLVR在复杂任务上的有效性。

Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation

👍 15

08/30 08:00

Sampled-token on-policy distillation (OPD) efficiently transfers capabilities from teacher to student using student-generated tokens, requiring teacher probabilities only for sampled tokens. Yet it frequently suffers from diversity distillation failure: the student's pass@1 improves while its pass@k

中文介绍 提出影响力导向蒸馏方法，解决采样token在线策略蒸馏中的多样性瓶颈。通过分析教师概率对多样性的影响，调整蒸馏目标，在提升学生pass@1的同时保持pass@k多样性。

WHALE: A Simple Recipe for Joint Harness-Weight Optimization

👍 13

08/31 08:00

Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness u

中文介绍 提出WHALE方法，联合优化智能体的模型参数和运行环境代码，避免单独优化导致的瓶颈。通过协同迭代更新权重和harness，实现更优的智能体性能，简单而有效。

Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

👍 13

09/01 08:00

This paper studies autonomous software development, in which LLM-based coding agents transform high-level requirements into complete, functional, and usable software systems without human intervention. We introduce Harness-of-Harness (HoH), a framework that enables coding agents to continually impro

中文介绍 提出Harness-of-Harness框架，支持多天自主软件开发并持续改进。智能体可生成完整软件系统，并通过对自身harness的反思与调整，实现代码质量的持续提升，减少人工干预。

VibeVoice-ASR-Streaming Technical Report

👍 10

09/02 08:00

Traditional speaker-attributed ASR systems treated ASR and speaker diarization as two separate tasks. Recently, end-to-end models such as VibeVoice-ASR have unified the two tasks within a single model. However, existing unified models still mainly support offline recognition, making it difficult to

中文介绍 报告VibeVoice-ASR-Streaming，将说话人属性ASR扩展到流式场景。在统一模型中同时处理语音识别和说话人分离，支持实时推理，便于在线交互应用，并讨论技术细节。

Agents in the Large: Perception-Centered Architecture for Persistent Agents

👍 9

08/31 08:00

Cognitive language agents have achieved substantial progress by equipping language models with memory, tools, and decision-making procedures, enabling agents to reason and act in interactive environments. Existing frameworks largely cast these agents as systems for solving user-specified, bounded ta

中文介绍 提出感知中心的持久智能体架构，使智能体在长时间、非有界任务中持续感知环境、积累记忆并自主决策，超越传统用户指定任务的局限，适用于现实世界的复杂交互。

CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing

👍 8

09/01 08:00

The attention prefilling phase of long-context LLM inference scales quadratically, making self-attention a severe computational bottleneck. Traditional sparse attention methods mitigate this through fixed patterns or offline profiling, but lack the flexibility to adapt to input-dependent attention s

中文介绍 提出CRISP，一种输入自适应稀疏预填充方法，通过结构质量驱动的路由，动态选择关键token进行注意力计算，降低长上下文预填充的二次复杂度，同时保持准确性。

Nobody is talking seriously about AI demand

@giovannicatt3 · 1.4K 粉丝 · 142.0K 阅 · 507 赞 · 56 转

09/01 23:51

AI capex for 2028 is forecast to be larger than the budget of France. Frontier AI labs’ revenue ramp justifies almost any number. Reflexivity in AI demand is a double-edged sword, and it's now a good

中文介绍 预测2028年AI资本支出将超过法国国家预算，前沿实验室收入增速支撑任何数字。作者以自反性为双刃剑，讨论AI需求可能过热的风险，提醒关注泡沫化。

Prediction: AI will collapse

@wordgrammer · 26.1K 粉丝 · 120.8K 阅 · 602 赞 · 33 转

09/03 05:30

The dot com bubble got some things right. It noticed that the internet was going to be big. It noticed that the value generated by the internet would roughly scale in proportion with the amount of

中文介绍 作者用互联网泡沫类比，认为AI泡沫将破裂，但网络价值量级判断正确。指出AI价值可能随算力规模增长，但当前估值远超实际，提醒警惕高估风险。

Compilers 2.0: AI as stochastic optimizer

@cdleary · 2.3K 粉丝 · 100.3K 阅 · 533 赞 · 66 转

09/02 04:00

There has been a lot of discussion following the presentation of the Jalapeño MLA kernel at HotChips and subsequent commentary by SemiAnalysis. As OpenAI’s hardware team, we just barely touched on

7 Grok Bots for Marketing

@irabukht · 18.4K 粉丝 · 63.3K 阅 · 536 赞 · 28 转

09/02 07:39

18 things to know and 7 paste-in prompts for running ads, SEO and GEO on Grok Bot. Collected from 1,000+ marketers in our community. Most of us use Grok Bot to run marketing work end to end: Weekly

中文介绍 汇总社区1000+营销人经验，提炼18条要点和7个可直接粘贴的Grok Bot提示词，覆盖广告投放、SEO和GEO，并分享每周工作流，帮助营销人员端到端使用机器人。

Introducing agentic video understanding with Gemini

@GoogleAIStudio · 198.5K 粉丝 · 37.4K 阅 · 522 赞 · 48 转

09/02 02:02

Our new agentic feature for video analysis cuts costs by up to 66% and reduces token consumption by up to 88% while boosting accuracy. Today, we’re launching agentic video understanding across our

中文介绍 Google推出Gemini智能体视频理解新功能，自称可降低66%成本和88%token消耗，同时提升准确率。该功能基于智能体工作流，已全面上线，强调效率与性能兼顾。

Using Grok Bot: 8 templates to get inspired

@mattyp · 57.4K 粉丝 · 29.7K 阅 · 508 赞 · 40 转

09/02 01:03

Templates are now available for Grok Bot, so I wanted to share my favorites. Templates contain the skills, memories, and official @bot plugins from your bots. When you use a template, you get a copy

中文介绍 Grok Bot现已支持模板功能，作者分享个人最爱的8个模板。模板包含技能、记忆和官方插件，复制即得全套配置，简化机器人搭建，适用于快速复用场景。

Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

@GoogleAIStudio · 198.5K 粉丝 · 26.2K 阅 · 507 赞 · 59 转

09/03 00:48

Our newest Gemini models deliver next-generation intelligence for agentic workflows and cybersecurity. Building on the momentum of 3.7 Flash from three weeks ago and marking our third Flash release in

中文介绍 Google发布Gemini 3.8 Flash及3.8 Flash Cyber，聚焦智能体与网络安全，距3.7 Flash仅三周，为三个月内第三个Flash版本，主打快速迭代、强化智能体工作流性能。

From rg to zg: Local Search Beyond Keywords

@QwenDevs · 10.9K 粉丝 · 21.4K 阅 · 509 赞 · 55 转

09/02 22:30

Summary: The information humans and agents need is often scattered across large numbers of local files, making it difficult to locate accurately and efficiently. zg (zvec-grep) is local-first search

中文介绍 Qwen团队推出zg（zvec-grep）本地优先搜索工具，从关键词转向向量语义。解决人和代理在海量本地文件中精确定位信息的需求，强调超越关键词匹配的准确性。

GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour

Latent Space · 09/04 05:09

We spent 20B+ tokens of GPT-6 Astra to explore everything. Here’s our learnings.

中文介绍 Latent Space 发文介绍 GPT-6 Astra，称其为自动化 AI 工程师，雇佣成本低于每小时 6 美元。团队用超过 200 亿 token 进行探索，并分享了相关经验与发现。

Introducing WeatherNext 3, our most advanced and accurate global weather AI model

DeepMind Blog · 09/03 23:02

中文介绍 DeepMind 推出 WeatherNext 3，称其为全球最先进、最准确的天气 AI 模型，旨在提升全球天气预报的准确性。

Daybreak for Frontline Defenders: $1B to protect essential services

OpenAI News · 09/03 21:15

OpenAI introduces Daybreak for Frontline Defenders. A $1 billion commitment expands access to frontier cyber AI, training, and support for essential services.

中文介绍 OpenAI 推出面向一线防御者的 Daybreak 计划，承诺投入 10 亿美元，扩大前沿网络 AI、培训和关键服务的支持范围。

NeoMME: an efficient Multimodal-native and Multilingual Encoder

Hugging Face Blog · 09/03 21:13

中文介绍 Hugging Face 发布 NeoMME，一种高效的多模态原生多语言编码器，旨在提升多语言和多模态处理效率。

Legora reviewed 41 documents in minutes with GPT-6 Astra

OpenAI News · 09/03 20:00

Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.

中文介绍 金融公司 Legora 使用 GPT-6 Astra 在几分钟内审阅 41 份文档，找出全部四个埋设错误，并在财务审查流程中绩效提升近 40%。

Playco cut manual fixes 50% prototyping games with GPT-6 Astra

OpenAI News · 09/03 20:00

Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.

中文介绍 游戏公司 Playco 使用 GPT-6 Astra 从灰色盒基础原型构建三个主题游戏原型，手动修复次数较前代模型减少 50%。

[AINews] Muse Spark 1.3 matches GPT-5.6-Sol, confirming Meta Superintelligence as the newest Frontier Lab, >90% discount for training

Latent Space · 09/03 12:38

an epic comeback story for Meta

中文介绍 Latent Space 报道称 Muse Spark 1.3 匹配 GPT-5.6-Sol 水平，Meta Superintelligence 成为最新前沿实验室，训练成本折扣超 90%，被视为 Meta 的史诗级回归。

Safety overview: GPT-6 Astra

OpenAI News · 09/03 08:00

GPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.

中文介绍 OpenAI 发布 GPT-6 Astra 安全概览，称其为最广泛部署的模型，也是首个达到其准备框架中网络安全能力关键级别的模型。

Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps

Hugging Face Blog · 09/03 08:00

中文介绍 Hugging Face 介绍通过 100 步 GRPO 微调 350M 参数模型，以改进结构化输出性能的方法。

Give Your Coding Agents a Memory You Own

Hugging Face Blog · 09/03 08:00

中文介绍 Hugging Face 文章探讨为编程智能体提供自主内存，以实现用户掌控的记忆管理。

Training a coding model to paint watercolours with TRL and OpenEnv

Hugging Face Blog · 09/03 08:00

中文介绍 Hugging Face 介绍使用 TRL 和 OpenEnv 训练编码模型绘画水彩画的方法。

Muse Spark 1.3 ✨, Gemini 3.8 Flash ⚡, intelligence vs cost 📊

TLDR AI · 09/03 08:00

中文介绍 TLDR AI 报道 Muse Spark 1.3、Gemini 3.8 Flash 等模型，并讨论智能与成本的权衡。

Proactive cyber defense for governments and enterprises

DeepMind Blog · 09/03 00:24

中文介绍 DeepMind 探讨为政府和企业的主动网络防御策略，强调预防性措施的重要性。

Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

DeepMind Blog · 09/03 00:18

中文介绍 DeepMind 推出 Gemini 3.8 Flash 和 3.8 Flash Cyber 模型，适用于通用及网络安全场景。

Facilitating AI integration with simplicity at scale

MIT Tech Review AI · 09/02 22:00

As companies scale, the technology supporting operations can become a liability just as quickly as it becomes an asset. Disconnected systems, site-specific tools, spreadsheets, and manual workarounds can create data silos that make it harder to spot problems early, coordinate responses, and make dec

中文介绍 MIT 科技评论文章指出，企业在规模扩大时，互联系统不足可能造成数据孤岛，需以简单方式集成 AI 以解决运营问题。